How do I account for the noise contribution of the image frequency in a single-sideband receiver?

In a single-sideband (SSB) receiver using a mixer for frequency down-conversion, the image frequency contributes noise that degrades the receiver's effective noise figure. A mixer responds to signals at both the desired RF frequency (f_RF = f_LO + f_IF) and the image frequency (f_image = f_LO - f_IF, or vice versa). While the desired signal comes from only one sideband, noise is received from both sidebands and folds into the IF band after mixing, effectively doubling the noise power compared to a receiver that rejects the image. The SSB noise figure of a mixer-based receiver is 3 dB higher than its double-sideband (DSB) noise figure: NF_SSB = NF_DSB + 3 dB. This is because in SSB operation, the signal comes from one sideband but noise comes from two. To reduce the image noise contribution: use an image-reject filter (a bandpass filter before the mixer that passes the desired RF band and rejects the image band; requires sufficient frequency separation between RF and image, at least 2 x f_IF; provides 20-40 dB of image rejection), use an image-reject mixer (IRC) or sideband-separating mixer (2SB) that uses two mixers with 90-degree phase offsets to separate the two sidebands into different IF outputs (provides 15-30 dB image rejection), choose a higher IF frequency to increase the IF separation between the RF and image bands (making the image filter easier to design), or use a direct-conversion (zero-IF) architecture where f_IF = 0 and the image frequency is the same as the RF frequency (no image problem but introduces DC offset and 1/f noise issues).